Neutralizing Antibody Evasion and Transduction with Purified Extracellular Vesicle-Enveloped Adeno-Associated Virus Vectors

Abstract

Adeno-associated virus (AAV) is classified as a nonenveloped DNA virus. However, several years ago, we discovered that in media of packaging cells producing recombinant AAV vectors, AAV capsids can associate with the interior and surface of extracellular vesicles (EVs), sometimes referred to as exosomes. Since then, we and others have demonstrated that exosome-enveloped AAV, exo-AAV, can enhance transduction in vivo as well as evade neutralizing antibodies.

While promising, these data were generated with differential centrifugation to pellet the exo-AAV. This method results in a heterogeneous mixture of exo-AAV, coprecipitating proteins, as well as free AAV capsids. To define the properties of exo-AAV more accurately, in this study, we used a density gradient method to purify exo-AAV. We next performed head-to-head comparisons of standard AAV1, differential centrifuged exo-AAV1, and gradient purified exo-AAV1 for antibody evasion and transgene expression in the murine brain. We found purified exo-AAV1 to be more resistant to neutralizing antibodies than the other AAV preparations. Direct intracranial injection of purified exo-AAV1 into mice resulted in robust transduction, which transduced a larger area of brain than standard AAV1. We also identified the recently described membrane-associated accessory protein by mass spectrometry of purified exo-AAV1 preparations. Finally, we used a scalable method, size-exclusion chromatography to isolate exo-AAV1, and demonstrated functional transduction in cultured cells and increased antibody resistance. Together, these data suggest that higher purity exo-AAV will have beneficial characteristics for gene delivery and also may lead to mechanistic insights into the incorporation of AAV into EVs.

INTRODUCTION

Viruses have traditionally been categorized as either having plasma- or organelle-derived membranes (enveloped virus) or those that exit the cell by a lytic mechanism and do not acquire a membrane (nonenveloped virus). However, the literature continues to grow that even classically defined nonenveloped viruses can acquire membranes, often in the form of extracellular vesicles (EVs). EVs are nano- to micron-sized lipid particles released by cells. They contain surface-exposed and luminal proteins, nucleic acids, and mediate cell-to-cell communication.¹ There are three main populations of EVs: exosomes (30–100 nm), which are derived from multivesicular bodies, budding vesicles commonly called microvesicles (∼100–1,000 nm), and apoptotic bodies that are released from dying cells (1–5 μm).² While there is some overlap between these EV subpopulations at both the size and biochemical levels, research continues to elucidate key differences in their biogenesis and functions. It seems that certain viruses may intentionally utilize EVs to enable their persistence. For example, hepatitis A virus, a nonenveloped picornavirus, can associate with exosomes to shield the capsid from neutralizing antibodies.³ Enteroviruses have been shown to pack multiple capsids inside EVs to allow “en-bloc” transmission to cells which increases infectivity.⁴

Several years ago, we made the surprising discovery that adeno-associated virus (AAV) vectors (a nonenveloped parvovirus) could be found on the surface and interior of EVs in the media of 293T producer cells.⁵ We hypothesized that this EV association would enable desirable features of AAV-based gene delivery. In subsequent studies, we showed that EV-associated AAV vectors, which we call exo-AAV, can enhance transduction of target cells in vivo,^6,7 and also evade neutralizing antibodies,^6,8 two clinically relevant features. To enable further development of the exo-AAV technology toward clinical application, several issues must be addressed. One such issue is testing the performance of a higher purity form of exo-AAV. It has been shown that media of AAV-producing 293T cells contains nonenveloped, “standard” AAV capsids,^9,10 and we have observed standard AAV capsids in electron micrographs of our exo-AAV preparations obtained by differential centrifugation.⁶ These coisolating standard AAV capsids could mask or confound interpretations of the functions of exo-AAV.

In this study, we asked what the effect of purifying exo-AAV and removing free standard capsids using a multistep density gradient would have on their antibody evasion and in vivo transduction profiles. We also assessed the presence of proteins in purified exo-AAV preparations by mass spectrometry. Finally, we tested a scalable method for purifying exo-AAV.

MATERIALS AND METHODS

All methods and experiments were approved by the Partners Institutional Biosafety Committee (IBC) of Partners Healthcare (Massachusetts General Hospital) under IBC approval 2011B000534.

Cells

Human 293T and HeLa cells were obtained from the American Type Culture Collection (Manassas, VA). These cells were cultured in high glucose Dulbecco's modified Eagle's medium (DMEM) (Life Technologies, Grand Island, NY) supplemented with 10% fetal bovine serum (FBS) (Sigma, St Louis, MO), 100 U/mL penicillin, and 100 μg/mL streptomycin (Life Technologies) in a humidified atmosphere supplemented with 5% CO₂ at 37°C.

Animals

All animal experiments were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care following guidelines set forth by the National Institutes of Health Guide for the Care and Use of Laboratory Animals. We used adult age (8–10 weeks old) C57BL/6 (strain # 000664) from The Jackson Laboratory (Bar Harbor, ME). Intracranially injected animals were euthanized 3 weeks postinjection, perfused transcardially with 4% formaldehyde in phosphate-buffered saline (PBS), and tissues were harvested and postfixed in 4% formaldehyde in PBS.

AAV vector production

For each production, we plated 15-cm tissue culture dishes with 1.5 × 10⁷ 293T cells/dish. The next day, cells were triple-plasmid transfected using the calcium phosphate method, with the adenovirus helper plasmid (pAdΔF6, 26 μg per plate), rep/cap plasmid (pXR1 for AAV1, 12 μg per plate), and ITR-flanked AAV transgene expression cassette (either single-stranded AAV-CBA-GFP or AAV-CBA-Fluc, 10 μg/plate, encoding green fluorescent protein or firefly luciferase, respectively) to induce production of AAV. The day after transfection, medium was changed to DMEM containing 2% FBS. Three days posttransfection, cell lysates (standard AAV) or cell culture media (exo-AAV) were harvested for further processing as described below.

Standard AAV cell lysate purification

AAV was purified from the freeze/thawed cell lysate using iodixanol density-gradient ultracentrifugation. Buffer exchange to PBS was done using ZEBA spin columns (7K MWCO; Thermo Fisher Scientific) and further concentration was performed using Amicon Ultra 100 kDa MWCO ultrafiltration centrifugal devices (Millipore). Vectors were stored at −80°C until use.

Differential centrifugation isolation of exo-AAV

exo-AAV was produced as previously described.¹¹ In brief, on days one and two after transfection, media was exchanged to DMEM containing 2% exosome-free FBS (made by overnight 100,000 g ultracentrifugation to deplete bovine exosomes). Media was harvested 3 days posttransfection and cell debris depleted by centrifugation at 300 g for 5 min and 1,000 g for 10 min. Next, large vesicles were pelleted by centrifugation at 20,000 g for 1 h and the media transferred to a new tube. This tube was centrifuged at 100,000 g for 1 h and the media was removed and exosome pellet containing exo-AAV was resuspended in 1 × PBS.

Isopycnic (density) gradient purification of exo-AAV

The cell culture media for purified exo-AAV was processed identically to the pelleted exo-AAV method up to and including the 20,000 g large vesicle depletion step. Next, the media was concentrated using tangential flow filtration (TFF) according to the manufacturer's instructions. In brief, a Masterflex L/S economy drive pump (Cole-Parmer, Vernon Hill, IL) fitted with an easy-load size 16 pump head (Cole-Parmer) was used to concentrate media through a VIVAFLOW 200, 10,000 molecular weight cutoff Hydrosart^® membrane cassette (Sartorius Stedim Biotech GmbH, Goettingen, Germany). Media was concentrated ∼10-fold (from 100 to 10 mL). Finally, media was filtered through a Sterile Syringe Filter, 0.8 μm Pore SFCA Membrane (Corning, Corning, NY). Media was treated with Benzonase (25 U/mL with 2 mM MgCl₂) for 1 h at 37°C. As previously described,⁸ we generated an 11-step gradient ranging from 8% to 60% iodixanol by hand using the overlay method. Two milliliter of concentrated, Benzonase-treated producer cell media was loaded on top of the gradient, which was loaded into a SW32Ti rotor (Beckman Coulter) and ultracentrifuged in an Optima XE-90 Ultracentrifuge (UC) (Beckman Coulter) at 175,000 g for 18 h at 4°C. At the end of the run, the rotor coasted to a stop with no brake. To isolate purified exo-AAV, we removed the ∼14–20% fraction (putative EV-containing fraction) corresponding to 10 mL of volume. In some experiments, we also collected the 40–60% layers containing EV-free AAV. Next, the exo-AAV containing fraction was subjected to buffer exchange (to PBS) and concentration using an Amicon Ultra 100 kDa molecular weight cutoff centrifugal device. Purified exo-AAV was pipetted into single-use aliquots and stored at −80°C until use.

Size-exclusion chromatography (SEC) column purification of exo-AAV

Cell culture media (15 mL) from one 15 cm plate of 293T (∼3 × 10⁷ cells) producing AAV vectors was harvested and media was clarified from cells debris and apoptotic bodies using sequential 300 g 5 min and 1,000 g 10 min spins. The supernatant was centrifuged for 60 min at 20,000 g, 4°C to deplete the larger microvesicles. The media was transferred to a new tube and treated with Benzonase (25 U/mL, Sigma) for 1 h at 37°C (MgCl₂ was added to final 2 mM concentration). Media was concentrated to ∼0.5 mL using 15 mL Amicon 100 kDa cutoff centrifugal device (Millipore). Size-exclusion column chromatography was performed using qEVoriginal (35 nm) columns attached to the Automatic Fraction Collector (both from Izon Science, Ltd., Medford, MA). We followed the exact manufacturer's instructions. In brief, 0.5 mL of the concentrated 293T-derived media was applied to the PBS-equilibrated column, and 12, 0.5 mL fractions were collected. EVs elute first in fraction 1–4 and smaller biomolecules, including free AAV, elute later. Fractions were subaliquoted and stored at −80° until use.

Vector quantitation (titers)

Before titration, exo-AAV and AAV samples were treated with DNase I to remove plasmid DNA from the transfection by mixing 5 μL of the sample with 1 μL DNase I, 5 μL 10 × buffer, and 39 μL water. Samples were incubated 1 h at 37°C and then DNase I was inactivated at 75°C for 15 min. We purified AAV genomes using High Pure Viral Nucleic Acid Kit (Roche, Indianapolis, IN) according to the manufacturer's protocol. Next, vector genome samples were titered (expressed as vector genomes/ml) using a quantitative TaqMan PCR that detects AAV genomes (polyA region of the transgene cassette) as previously described.¹²

Anti-AAV neutralization assay

In vitro neutralization assays were performed with Gamunex-C purified intravenous immunoglobulin (IVIg), (Grifols, Barcelona, Spain) or with pooled normal human sera (Innovative Research, Novi, MI). HeLa cells were seeded at 10,000 cells per well in a 96-well plate the day before the assay. Next, a dose of 1 × 10⁵ vg/cell of AAV1-FLuc or exo-AAV1-Fluc was mixed with serial dilutions of human serum or IVIg in FBS-free media. Vector samples mixed with media containing no human serum served as control. These were incubated for 1 h at 37°C before adding to cells for 1.5 h at 37°C. After washing cells one time, and replacing with complete medium, cells were incubated for 48 h before performing a luciferase assay using Bright-Glo™ Luciferase reagent (Promega, Madison, WI). A BioTek Synergy HTX multimode luminometer (BioTex, Winooski, VT) was used to detect luminescence. Luciferase values for each sample, expressed in relative light units (RLUs), were plotted as a fraction of the AAV transduction sample without serum (which was set to 1.0). In one set of experiments, we used vectors encoding GFP (AAV-CBA-GFP) instead of FLuc. This experiment was performed similarly to the above using HeLa cells and IVIg, except at 48 h posttransduction, cells were harvested for flow cytometry using a BD LSRFortessa™ flow cytometer (BD Biosciences, Franklin Lakes, NJ) at the MGH Flow and Mass Cytometry Core. After forward and side scatter parameters were set, and singlets were gated on, we measured the percentages of GFP-positive cells, using nontransduced cells to set the threshold of GFP positivity (BD FACSDiva Software, BD Biosciences).

Intracranial injection of AAV vectors

Adult C57BL/6 mice (n = 3/group) were injected stereotactically into the striatum with 3.2 × 10⁸ vg of each vector preparation in a volume of 2 μL. Mice were placed into a Just For Mouse Stereotaxic Frame (Stoelting, Wood Dale, IL). AAV vectors were infused in the left midstriatum using the following coordinates from bregma in mm: AP +0.5, ML +2.0, DV −2.5. AAV vectors were infused into the striatum at a rate of 0.2 μL/min using a Quintessential Stereotactic Injector pump (Stoelting) to drive a gas-tight Hamilton Syringe (Hamilton, NV) attached to a 10 μL 33-gauge NEUROS model syringe (Hamilton, NV).

Tissue processing and immunofluorescence imaging

Brains from mice stereotactically injected with vectors encoding GFP were postfixed in 4% formaldehyde diluted in PBS for 48 h, followed by 30% sucrose for cryopreservation for another 48–72 h, after which brains were embedded and frozen in Tissue-Tek^® O.C.T. compound (Sakura Finetek USA, Torrance, CA). Coronal floating sections (40 μm) were cut using a NX50 CryoStar Cryostat (Thermo Scientific). After rinsing off the sucrose in PBS, cryosections were permeabilized with 0.5% Triton™ X-100 (Millipore Sigma) in PBS for 30 min at room temperature and blocked with 5% normal goat serum (or normal donkey serum) and 0.3% Triton in PBS for 1 h at room temperature. Primary antibodies were incubated 3 days at 4°C in 1.5% normal goat serum and 0.3% Triton in PBS, while Alexa Fluor 488, Alexa Fluor 555 (Thermo Scientific), or Cy3 conjugated secondary antibodies (Jackson ImmunoResearch laboratories) were incubated for 1 h the next day. Primary antibodies used for this study were Mouse anti-NeuN (Abcam), rabbit anti-Olig2 (EMD Millipore); rabbit anti-Iba1 (Wako, Japan); rabbit anti-GFAP (Dako); and rabbit anti-GFP (Thermo Scientific). Sections were mounted with VECTASHIELD mounting medium with DAPI (Vector Laboratories, Burlingame, CA). Imaging was performed with Zeiss confocal microscope.

To quantitate the percentages of GFP-positive cells, which were NeuN-positive neurons or GFAP-positive astrocytes, we performed manual counting of high magnification images of the cortex or caudate putamen on the ipsilateral hemisphere. The total number of GFP-positive cells were counted (38–102 cells/image counted) and the number of colabeled marker (either NeuN or GFAP) was then counted. The ratio of colocalized GFP-positive cells/total GFP-positive cells was used to calculate the tropism for the given cell type. For calculation of transduction efficiency, the total number of NeuN- or GFAP-positive cells were counted in each image (30–117 cells/image counted) and the number of GFP-positive cells that colocalized with these markers were then counted. The ratio of the number of GFP⁺ cell marker positive cells/total cell marker positive cells was used to calculate the transduction efficiency.

Fluorescence imaging of whole brain coronal sections was performed using a KEYENCE BZ-X800 microscope (KEYENCE Corporation of America, Itasca, IL). A 4X overview of each slide was taken before scanned 10X images of the coronal brain section were acquired. Images were automatically stitched using BZ-X100 Analyzer Software (KEYENCE).

For quantitation of relative transduction area spanning the injection site, we used a fluorescence micrograph which consists of all three vector groups with three mice/group and five coronal sections/mouse. This provided 15 images/vector group to analyze for peak transduction area. This image (without sample identifiers) was analyzed with ImageJ (Version 1.50b, National Institutes of Health¹³) by a researcher who did not know the identity of the samples. Using the freehand selection tool, regions of interest were carefully drawn around the GFP-positive area in each section and the “Analyze-Measure” function was used in ImageJ to calculate relative area. Next, we normalized all values to the lowest sample, which was arbitrarily set to 1.0.

Sodium dodecyl sulfate polyacrylamide gel electrophoresis and immunoblots

Samples from gradient purified exo-AAV, UC exo-AAV, SEC fractions, or standard AAV were loaded onto a 4–12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) Bis-Tris gel (Thermo Fisher Scientific) and electrophoresed for 1.5 h at 120 V. Proteins were transferred to nitrocellulose membrane for 1 h at 30 V. For detection of AAV capsid proteins VP1, VP2, and VP3, rabbit polyclonal antibody to AAV capsid was used (product #03-61084, American Research Products, Waltham, MA). For detection of EV protein, CD9, we used anti-human CD9 antibody (D801A) (cat. no. 13174, Cell Signaling Technology, Danvers, MA). For detection of EV protein CD81, we used an anti-CD81 antibody (D3N2D) (cat. no. 56039, Cell Signaling Technology). A secondary horse radish peroxidase-conjugated anti-rabbit antibody (cat. no. NA934VS, Cytiva, Marlborough, MA) was used for detection of both, capsid, CD81 and CD9. Chemiluminescence was detected with SuperSignal™ West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) and membranes exposed to HyBlot CL autoradiography film (labForce, Thomas Scientific).

Liquid chromatography-mass spectrometry of AAV and exo-AAV preparations

Mass spectrometry was performed at the ZMNH Core Facility for Mass Spectrometry & Proteomics at the Center for Molecular Biology at the University of Heidelberg. In brief, standard AAV1 or gradient-purified exo-AAV1 preparations were loaded onto an SDS-PAGE gel and electrophoresed. The Coomassie-stained bands were cut out and processed as described with minor modifications.¹⁴ In brief, trypsin digestion was done over night at 37°C. The reaction was quenched by addition of 20 μL of 0.1% trifluoroacetic acid (TFA; Biosolve, Valkenswaard, The Netherlands) and the supernatant was dried in a vacuum concentrator before LC-MS analysis. Nanoflow LC-MS² analysis was performed with an Ultimate 3000 liquid chromatography system coupled to an QExactive HF mass spectrometer (Thermo-Fischer, Bremen, Germany). Samples were dissolved in 0.1% TFA, injected to a self-packed analytical column (75 μm × 200 mm; ReproSil Pur 120 C18-AQ; Dr Maisch GmbH), and eluted with a flow rate of 300 nL/min in an acetonitrile-gradient (3–40%). The mass spectrometer was operated in data-dependent acquisition mode, automatically switching between MS and MS². Collision-induced dissociation MS² spectra were generated for up to 20 precursors with normalized collision energy of 29%.

Database search of LS-MS data

Raw files were processed using MaxQuant version 1.6.12.0¹⁵ for peptide identification and quantification. MS² spectra were searched against the UniProt human proteome database (UP000005640_9606.fasta downloaded November 2019), the contaminants database provided together with the MaxQuant software and a database containing the sequence of exo-AAV using the Andromeda search engine with the following parameters: Carbamidomethylation of cysteine residues as fixed modification and Acetyl (Protein N-term), Oxidation (M) and deamidation (Q,N) as variable modifications, trypsin/P as the proteolytic enzyme with up to two missed cleavages was allowed. The maximum false discovery rate for proteins and peptides was 0.01 and a minimum peptide length of seven amino acids was required. All other parameters were default parameters of MaxQuant. Quantitative normalized ratios were calculated by MaxQuant and used for further data analysis.

Immunocapture and characterization of EVs

To quantitate EV quantity, tetraspanin marker expression, and size we used the ExoView^® R100 instrument in conjunction with the ExoView Tetraspanin Kit (NanoView Biosciences, Brighton, MA). In brief, this instrument consists of a fluorescence and interferometric microscope, which detects single EVs captured on chips coated with antibodies specific to tetraspanin proteins expressed on EVs. The instrument can size EVs (range 50–200 nm) and also quantitate numbers of EVs by detecting fluorescence from bound labeled antibodies. Fractions from SEC were diluted 1:400 and incubated 16 h at room temperature on tetraspanin chips printed with antibodies against CD81, CD63, CD9 or a isotype control antibody. The following day, the chips were rinsed four times with rapid shaking on an orbital shaker. Next, fluorescently labeled secondary antibodies specific for CD9, CD63, and CD81 were added to the chip and incubated for 1 h. After extensive washing, chips were dried and then inserted into the ExoView instrument for scanning and data acquisition. Data were analyzed using the ExoView Analyzer software.

Cryo electron microscopy

Analysis of SEC fraction 1 was performed by UMass Medical School Cryo-EM Core Facility. Frozen sample at −80°C was sent to the facility. Samples were thawed, 3 μL placed on the glow-discharged R1.2/1.3 C-flat holey carbon copper grids, back-blotted with filter paper, and frozen in a liquefied ethane/propane mixture using an EMS-2 rapid immersion freezer. Grids were stored in liquid nitrogen until imaging. Images were recorded at a defocus −2.5 μm on a Thermo Fisher Scientific Talos Arctica electron microscope operating at 200 kV with a K3 direct detection camera (Gatan) in counting mode with pixel size at 2.27 Å (Magnification 17.5k). Images were collected with a total electron dose of 39.08 e-/Å2 for each image under a low-dose mode using the SerialEM software.¹⁶

Statistics

All data were graphed and statistical analysis performed using GraphPad Prism 9.0.2 software (San Diego, CA). All data are reported as mean +/− standard deviation. A p-value of <0.05 was considered statistically significant. Comparison of transduction area across vector groups was performed using a one-way analysis of variance (ANOVA) followed by a Dunnett's multiple comparisons test. Comparison of transduction efficiency between vector groups across different IVIg concentrations was performed using a two-way ANOVA followed by a Dunnett's multiple comparisons test.

RESULTS

Purification of AAV and exo-AAV vectors and biochemical characterization

The overall goal of this study was to assess properties of highly purified exo-AAV in comparison to conventional nonenveloped AAV and with exo-AAV generated by a pelleting method that we have utilized in several publications.^{5–8,11,17,18} We hypothesized that removal of free capsids and other coprecipitating proteins found in the UC pelleted exo-AAV preparations would improve resistance to anti-AAV antibody-mediated neutralization. Figure 1a depicts the different methods of isolation of each preparation of AAV vectors for this study.

Figure 1.

Overview of processing and purification of standard AAV and exo-AAV. (a) AAV vectors encoding GFP or FLuc are produced by transient transfection of plasmids in 293T cells. Cell lysate-derived AAV is purified on a traditional iodixanol gradient to isolate nonenveloped AAV from the 40% layer. The conditioned media is split into two fractions, of which the first is used to pellet exo-AAV using ultracentrifugation. This results in some copelleting of nonenveloped AAV and other biomolecules. The second media fraction is concentrated by tangential flow filtration and loaded on an 11-step iodixanol density gradient. This allows purification of exo-AAV in the 14–20% layers and separation from nonenveloped AAV in the 40–60% (denser) layers. SEC purification of exo-AAV was also tested from the second fraction in some experiments. (b) Overview of experiments. Conventional nonenveloped AAV, pelleted and resuspended exo-AAV, and gradient or SEC-purified exo-AAV are compared for their ability to evade neutralization by anti-AAV antibodies and mediate transduction in vitro and in vivo. We also performed mass spectrometry on some proteins in the preparations. AAV, adeno-associated virus; SEC, size-exclusion chromatography. Color images are available online.

Standard nonenveloped AAV was isolated from 293T cell lysates and purified by iodixanol density gradient ultracentrifugation. To isolate UC-pelleted exo-AAV, we subjected conditioned media to sequential centrifugation and ultracentrifugation steps and resuspended the pelleted material in PBS. To isolate purified exo-AAV, we concentrated conditioned media using TFF and then loaded it onto an 11-layer iodixanol step gradient ranging from 8% to 60% iodixanol. We⁸ and others¹⁹ have previously demonstrated that this gradient separates less dense exo-AAV (14–20% fractions) from denser free AAV capsids (40–60% fractions).

Next, we used standard AAV1-GFP, UC-pelleted exo-AAV1-GFP, and gradient-purified exo-AAV1-GFP in biochemical characterization assays, in vitro neutralization assays, and in vivo gene delivery experiments (Fig. 1b). First, we performed nanoparticle tracking analysis (NTA) to look at the EV particle concentration and size distribution in the gradient-purified preparation of exo-AAV. The total EV yield as measured in total particles by NTA was 4.4 × 10¹¹ particles. The size profile demonstrated a peak at 128 nm with a shoulder at 185.5 nm (Fig. 2a). Next, we analyzed the vector preparations by looking at the capsid proteins of AAV (VP1, 2, 3) and the EV marker, CD9 by immunoblot (Fig. 2b). In addition to the 14–20% fraction containing the putative exo-AAV vector, we analyzed the 40–60% fraction from the same gradient, which should contain free AAV. As expected, the cell lysate-purified AAV did not contain detectable CD9, while VP bands were readily detected. In contrast, in vector isolated from conditioned media, we detected both VP bands as well as CD9 in both the UC-pelleted exo-AAV as well as the gradient-purified exo-AAV (14–20% fraction). As predicted, the 40–60% fraction from the media gradient contained free AAV capsids, and CD9 was undetectable. Interestingly, we noticed a pronounced shift in VP3 from purified exo-AAV, which migrated at a lower molecular weight compared to all other preparations/fractions (Fig. 2b). This same shift was observed in independent preparations of gradient purified exo-AAV (data not shown).

Figure 2.

Biochemical characterization of preps. (a) Size and concentration profile of the gradient purified exo-AAV fraction determined by nanoparticle tracking analysis (b). Each lane of an SDS-PAGE gel was loaded with equal vg of (1) CL-derived, iodixanol gradient-purified AAV1, or media-derived (2) pelleted exo-AAV, or (3) exo-AAV and free AAV isolated from the 14% to 20% fraction and 40% to 60% fraction of a multistep iodixanol gradient, respectively. The top immunoblot is for anticapsid proteins, VP1, VP2, and VP3, while the bottom immunoblot is for the EV marker, CD9. (c) Coomassie-stained SDS-PAGE gel showing bands of CL-derived, iodixanol-purified standard AAV1, and media-derived, purified exo-AAV1 isolated from the 14% to 20% iodixanol gradient fraction. Red boxes indicate bands excised for mass spectrometry analysis. L, ladder. Approximate migration pattern of VP1, VP2, and VP3 is indicated by text. Mass spectrometry fingerprint data were entered into the Mascot search engine to identify proteins. The values in the boxes indicate iBAQ (intensity based absolute quantification) values. CL, cell lysate; EV, extracellular vesicle; SDS-PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis. Color images are available online.

To characterize the capsid and other proteins associated with purified exo-AAV, we gel-extracted the putative VP3 band from either standard AAV1 or purified exo-AAV1 resolved on SDS-PAGE gels stained with Coomassie dye (we called this band #1). We also notice a small band migrating at ∼12 kDa in the exo-AAV1 lane, which was not observed in the standard AAV lane. We gel-extracted this small molecular weight protein band (band #2) and the corresponding region of the standard AAV1 lane in case the protein was at a level undetectable by Coomassie staining (Fig. 2c). The gel-extracted VP bands and the small molecular weight bands for each preparation of AAV were submitted to mass spectrometry. The VP3 bands (band #1) of standard and exo-AAV1 did not show differences in the detected peptide composition of AAV VP (Fig. 2c). Interestingly, we could detect the recently discovered AAV-encoded protein membrane associated accessory protein (MAAP²⁰) in the small molecular weight band #2 of exo-AAV1, but not in the corresponding region of the gel for standard AAV1.

Increasing the purity and removal of nonenveloped AAV increases resistance of exo-AAV to antibody neutralization

To test whether increasing the purity and removal of free AAV would affect antibody-mediated neutralization, we performed head-to-head comparison of standard AAV1, UC-pelleted exo-AAV1, and purified exo-AAV incubated in the presence of media, or three different concentrations of IVIg, which contains anti-AAV antibodies. Three posttransduction of HeLa cells were processed and subjected to flow cytometry to determine the percentage of GFP-positive cells in each sample.

In the absence of IVIg, all vector preparations transduced similar levels of cells with means ranging from 9.2% to 14% (Fig. 3a). For standard AAV, concentrations of IVIg at 0.1, 0.2, and 0.4 mg/mL reduced transduction by 48%, 70%, and 91% compared to no IVIg, respectively (Fig. 3a). A similar trend was observed with UC-pelleted exo-AAV1. Remarkably, purified exo-AAV1 transduction was minimally impacted by 0.1 or 0.2 mg/mL of IVIg, with transduction being reduced by only 17% (0.1 mg/mL) and 21% (0.2 mg/mL) compared to no IVIg (Fig. 3a). Only at the 0.4 mg/mL concentration did transduction by purified exo-AAV1 drop by greater than half (60%) compared to no IVIg. We performed a two-way ANOVA to compare how vector group and IVIg concentration affected transduction efficiency. We found that purified exo-AAV was statistically significantly more resistant to IVIg at all concentrations compared to both standard AAV and pelleted exo-AAV (Fig. 3b, p < 0.0001).

Figure 3.

Gradient-purified exo-AAV1 resists antibody neutralization better than standard AA1V and ultracentrifugation-pelleted exo-AAV1. Standard AAV1, pelleted exo-AAV1, and purified exo-AAV1 (10⁵ vg/cell) were incubated with media alone or media containing 0.1, 0.2, and 0.4 mg/mL concentrations of IVIg. Next, complexes were added to HeLa cells and transduction (% GFP⁺ cells) was measured 3 days later by flow cytometry. (a) Percentage of GFP-positive cells for each vector preparation/condition. (b) For each of the vector preparations, the values for the 0.1, 0.2, and 0.4 mg/mL IVIg concentrations were normalized to the mean of the vector's “no IVIg” sample, which was arbitrarily set to 1.0. Error bars show standard deviation of the mean. ****p < 0.0001. IVIg, intravenous immunoglobulin. Color images are available online.

Transduction profile of purified exo-AAV1 after direct intracranial injection

We next set out to test the transduction profile of gradient-purified exo-AAV1 compared to standard cell lysate-purified AAV1 and UC-pelleted exo-AAV1 in vivo. Adult C57BL/6 mice (n = 3/group) were injected stereotactically into the striatum with 3.2 × 10⁸ vg of each vector preparation in a volume of 2 μL. Three weeks later, mice were killed, and brains coronally cryosectioned spanning the injection site. We examined whole coronal sections at low magnification by fluorescence microscopy of intrinsic GFP fluorescence (Fig. 4a). All mice in each group showed bright GFP fluorescence on the ipsilateral hemisphere. In both, standard AAV1-GFP- and UC-exo-AAV1-GFP-injected mice, transduction was restricted near the injection site. Interestingly, in 2 of 3 purified exo-AAV1-GFP-injected mice, GFP expression penetrated toward and into the contralateral hemisphere as seen by GFP expression in corpus callosum, lateral septal nucleus, and cingulum. Image software quantitation of the relative GFP-positive area in sections immediately surrounding the injection site demonstrated a significant (p = 0.0009) increase in transduction area with purified exo-AAV1 compared to standard AAV1. Figure 4b suggests greater spread from the injection site with this preparation of vector.

Figure 4.

Gradient-purified exo-AAV1 mediates robust brain transduction after intracranial injection in adult mice. Adult C57BL/6 male mice were intracranially injected in the left hemisphere with standard AAV1-GFP, UC-exo-AAV1-GFP, or gradient-purified exo-AAV1-GFP, each with 3.2 × 10⁸ vg (n = 3 mice/group). Animals were killed 4 weeks postvector injection. (a) Whole coronal section scans showing intrinsic GFP fluorescence in the injected hemisphere of all three groups. Depicted are serial sections around the injection site from all three mice per group. (b) Quantitation of the relative transduced area in each section in (a) surrounding the injection site. Bold dotted line on violin plot indicates median value for each group. ***p = 0.0009. ns, not significant. (c) High magnification images of GFP-positive neurons (costained with NeuN) in the cortex and caudate putamen in mice injected with the three vector preparations. Scale bar = 20 μm. UC, ultracentrifuge. Color images are available online.

Next, we determined the phenotype of transduced cells using colabeling for neurons (NeuN), astrocytes (GFAP), microglia (Iba1), and oligodendrocytes (olig2). In the injected hemisphere, the majority of transduced cells for all vector types were neurons (Fig. 4c and Table 1). This was consistent in both, the cortex and caudate putamen. In the contralateral hemisphere, only rare GFP, positive cells were detected (data not shown). Interestingly, there were less GFP⁺ cells that colocalized with NeuN for purified exo-AAV1 compared to standard AAV1 (Table 1). Further transduction efficiency of neurons and astrocytes was slightly lower for purified exo-AAV1 compared to standard AAV1 (Table 1).

Table 1.

Quantitation of cell type specificity and efficiency after intracranial injection

	Cell Type Tropism				Transduction Efficiency
	%GFP+, NeuN+ Cells		%GFP+, GFAP+ Cells		%GFP Positive Neurons (NeuN)		%GFP Positive Astrocytes (GFAP)
Vector Type	Cortex	CP	Cortex	CP	Cortex	CP	Cortex	CP
AAV1	69.6%	60.3%	31.4%	14.7%	66.5%	65.4%	43.8%	31.3%
UC exo-AAV1	ND	50.4%	26.3%	18.2%	ND	38.4%	30%	19.6%
Purified exo-AAV1	27.2%	33.7%	33.9	17.4%	51.1%	46.3%	21.9%	19.4%

AAV, adeno-associated virus; CP, caudate putamen; ND, not determined; UC, ultracentrifuge.

Transduction of astrocytes was detectable for all vector preparations, although to a much lesser degree than neurons (Supplementary Fig. S1). GFAP⁺/GFP⁺ cells were mainly restricted to the cortex and cells lining the ventricles. For standard AAV1, we observed no colocalization of GFP with microglia or oligodendrocytes. In contrast, rare GFP-positive microglia and oligodendrocytes were detected in UC-exo-AAV1 and purified exo-AAV1-treated groups, respectively (Supplementary Fig. S2).

Scalable purification of exo-AAV

The iodixanol-purified exo-AAV1 vector displayed enhanced AAV antibody evasion properties and at least equally potent transduction in vivo compared to standard AAV1 capsids. This gave us rationale that highly purified exo-AAV via removal of free AAV capsids could provide a promising vector for clinical development. However, density gradients are not very scalable; therefore, we next sought to develop a scalable method of purification of exo-AAV. To this end, we tested a commercially available SEC column with a 35-nanometer pore size, which was developed for purification of EVs (Fig. 5a). EVs are recovered in fractions 1–4, and we reasoned that exo-AAV would be enriched in these fractions, with smaller nonenveloped AAV eluting in later fractions. We harvested media from 293T cells (15 mL, 3 × 10⁷ cells) producing AAV1-FLuc vectors and concentrated this media to 0.5 mL and loaded it onto the SEC column. Fractions 1–12 were recovered and titered for AAV genomes. We found that ∼42% of AAV1 genomes on the gradient were recovered in fractions 1–4 (Fig. 5b). The yields and production information from independent SEC runs of conditioned media from AAV1 and AAV9 producing 293T cells are displayed in Table 2. In both runs a large percentage of AAV genomes were obtained in the EV-rich fractions (42–61%). To better understand if the putative EV fractions (1–4) contain bona-fide EV-associated AAV, we treated some AAV1-Fluc-containing media with the detergent Igepal. Treatment with detergent will lyse many of the vesicles and release free AAV, which should shift to the later, nonenveloped AAV fractions. As predicted, Igepal treatment decreased the fraction of AAV genomes in the putative EV fractions (compared to untreated media) and increased the percentage in the nonenveloped AAV fractions (Fig. 5b, c). There was 37% of genomes that were “detergent-sensitive” in fractions 1–4, which suggests that at least this amount (some EVs may have been resistant to lysis as reported²¹) was bona fide exo-AAV. We confirmed that the AAV capsid proteins VP1–3 were detected in fractions 2–4 (Fig. 5d). We also confirmed the EV-associated tetraspanin protein, CD81, was readily detected in SEC fractions 1–4, but not 8, 9, or with standard AAV1 (Fig. 5e). We next performed transduction and neutralization assays on HeLa cells on putative exo-AAV fractions 2–4 compared to standard AAV1. Transduction levels (in RLUs) were similar for fractions 1–4 while they were slightly, yet, statistically significantly higher for AAV1-transduced cells (Fig. 5f). We also performed a neutralization assay with IVIg at six concentrations (0.05, 0.1, 0.2, 0.4, 1.0, and 2.0 mg/mL). We observed a strikingly higher resistance of all fractions compared to standard AAV1. In particular, SEC fraction 3 had a highly significant (p < 0.0001) difference compared to standard AAV1 at all but the two highest concentrations (1.0 and 2.0 mg/mL) of IVIg (Fig. 5g). We also compared the ability of IVIg to neutralize a later SEC fraction, #8, which should contain mainly nonenveloped AAV capsids (as indicated by low CD81 detection by western blot, Fig. 5e). To do this, we performed an independent experiment and isolated fractions on the SEC column. We compared SEC fractions 2 (an exo-AAV fraction) and standard AAV1 to fraction 8 at 4 IVIg concentrations (0.05, 0.1, 0.2, and 0.4 mg/mL). As predicted, SEC fraction 8 had a similar neutralization profile to standard AAV1, while SEC fraction 2 had a highly significant increased resistance to IVIg neutralization than did SEC fraction 8 (Fig. 5h). To further characterize the SEC fractions, we designed an experiment to answer two questions. The first was to assess whether, similar to IVIg, would exo-AAV isolated by SEC have resistance to pooled normal human sera? The second question addressed whether “plain” 293T-derived EVs isolated by the same SEC method had an ability to protect standard nonenveloped AAV from neutralization. We performed two neutralization assays using either IVIg or pooled human sera and the following test samples: standard AAV1, exo-AAV1 from SEC fraction 1, and standard AAV1 spiked with EVs from 293T (no AAV vector produced) isolated from SEC fraction 1. As expected, exo-AAV1 from SEC fraction 1 had higher resistance to IVIg than standard AAV1 (Supplementary Fig. S3a). The same finding was observed for pooled human sera (Supplementary Fig. S3b). For both neutralization assays, there was minimal impact of EVs on standard AAV1 neutralization (Supplementary Fig. S3a, b). Put a different way, 293T-derived EVs did not greatly enhance vector resistance to neutralization to the level that exo-AAV1 does.

Figure 5.

Scalable isolation of exo-AAV1 using SEC with a 35 nm pore size resin. (a) Concentrated media from 293T cells producing AAV1-FLuc vectors was applied to the SEC column (0.5 mL applied volume) with the goal of separating larger exo-AAV from free AAV capsids. (b) Fractions of 0.5 mL were collected and subjected to qPCR to detect vector genomes (black line and circles). A separate column was run with media treated with 2% Igepal to confirm the presence of detergent-sensitive exo-AAVs in fractions 1–4 (orange line and squares). The putative EV fractions 1–4 are indicated. (c) Detergent treatment decreases AAV genomes in fractions 1–4 and increases AAV genomes in fractions 5–12. (d) Western blot analysis of selected SEC fractions #1–4 and #8. The membrane depicts electrophoresed SEC fractions and standard AAV (equal vg loaded/sample) probed with an anticapsid antibody. (e) Western blot detection of the tetraspanin protein CD81 in EV-rich fractions 1–4. (f) Transduction of HeLa cells by SEC fractions 1–4 compared to standard AAV1-FLuc. Equal doses (vg/cell) of each vector were added per well. Error bars indicate standard deviation of the mean. **p = 0.0012, ****p < 0.0001. (g) Neutralization assay using IVIg. Transduction at each concentration of IVIg was normalized to the transduction in the absence of IVIg, which was arbitrarily set at 1.0. Error bars indicate standard deviation of the mean. ****p < 0.0001, fraction 3 compared to standard AAV at 0.05, 0.1, 0.2, and 0.4 mg/mL IVIg. (h) Neutralization assay using IVIg as in (g) but comparing SEC fraction 1 versus fraction 8 and standard AAV1. Fraction 8 versus fraction 2: ****p < 0.0001; ***p < 0.0007. Color images are available online.

Table 2.

Yield information from size-exclusion chromatography fractions

Vector	Yield EV Fractions 1–4 (vg)	Yield Fractions 5–12 (vg)	Percent of Genomes in EV Fraction
AAV1-FLuc	1.40 × 1011	1.93 × 1011	42
AAV9-FLuc	2.27 × 1011	1.46 × 1011	61

EV, extracellular vesicle.

To further characterize the putative exo-AAV fractions, we used the ExoView system, which consists of a chip that captures single EVs using specific antibodies to tetraspanins. We found in pilot studies that immunocapture of 293T-derived EVs with the tetraspanin CD81 was very efficient, so we used this capture marker for analysis. Sizing information of CD81-captured EVs was obtained by interferometric microscopy and single particle detection by fluorescently-tagged secondary antibodies to the EVs. We found the mean sizes of fractions 1–4 to range from 59 to 65 nm (Fig. 6a). As expected, the larger mean EV sizes were in the earlier fraction 1. Next, we determined the total number of CD81-captured particles for SEC fractions 1–4. Fractions 1 and 2 had the highest particle amount, with an average of 9.74 × 10⁹ CD81-captured particles for fraction 2 (Fig. 6b). Since we calculated both the CD81-captured particle number as well as total AAV genomes in each fraction, we were able to calculate the ratio of AAV VGs/particle in fractions 1–4. There was a mean range of 4.33 (fraction 2) to 14.8 (fraction 3) VG/particle (Fig. 6c). We also examined SEC fraction 1 by cryo-electron microscopy to look at vesicle size and morphology. We observed several EV structures, including single vesicles as well as vesicles within vesicles, with size ranging from ∼50 to over 200 nm (Supplementary Fig. S4). We were able to detect AAV capsids within smaller and larger EVs, as well as nonenveloped (EV-free) forms of AAV (Fig. 6d and Supplementary Fig. S4).

Figure 6.

Size, quantitation, and morphology of SEC fractions containing exo-AAV1. (a) Average sizes of CD81-captured EVs as measured by interferometric microscopy. *p = 0.0285; ****p = 0.0001. (b) Yield of CD81-captured EVs from each fraction isolated from one 15 cm plate of 293T cells producing AAV1-FLuc. **p < 0.01, ***p = 0.0004, ****p < 0.0001. (c) Ratio of AAV vector genomes (calculated by qPCR) to CD81-captured particles for each SEC fraction. (d) Detection of icosahedral AAV capsids of ∼20–25 nm inside an EV (arrowheads point to capsids). A contrast-enhanced image is on the right to more readily visualize capsids. Scale bar = 100 nm. Color images are available online.

DISCUSSION

In this work, we set out to characterize whether purified exo-AAV could transduce cells and resist neutralizing anti-AAV antibodies. The premise of this study is founded on prior work from our laboratory^5–8,11,18 and others^19,22–24 showing that exosome-associated AAV (exo-AAV) vectors enhance transduction of cells in vivo and can also resist antibody neutralization. While these results were promising, they were most often performed using a differential centrifugation process that (1) results in coisolation of free AAV capsids, as well as other proteins in addition to EVs and (2) was not a scalable method. To first test whether we could separate exo-AAV from free AAV capsids in culture media, we used a multistep iodixanol gradient that has been used by ourselves⁸ and others^3,19 for purification of EV-associated virus particles. We harvested exo-AAV from the less-dense 14–20% iodixanol fractions and free AAV was found in the 40–60% layers as expected. We performed a direct comparison of standard AAV1, UC-exo-AAV, and the density gradient-purified exo-AAV. As predicted, the gradient-purified exo-AAV1 had the highest resistance to neutralization. These data are important, as they suggest that the resistance to neutralization is not from “decoy” free capsid in the preparations, but due to the properties of EV-associated AAV. UC-exo-AAV1 neutralized in a similar manner to standard AAV1. We have previously reported that UC exo-AAV8 and UC exo-AAV9 vectors have higher resistance to neutralization than standard AAV8 and AAV9 serotypes.^6,8 The apparent discrepancy for enhanced resistance with UC exo-AAV8 and 9 vectors compared to UC exo-AAV1 vector may be due to higher amounts of free AAV1 capsids or AAV1 capsids on the outside of EVs in the latter preparation. In any case, our data show that increasing the purity of exo-AAV1 increases its antibody resistance.

We next asked whether we could purify exo-AAV from media using a scalable method. We chose SEC as it should allow separation of larger exo-AAV particles from free AAV capsids without the need for affinity methods. The latter relies on discrimination between EVs and AAV capsids for ligands, which may be challenging. We tested a commercially available SEC column developed for isolating EVs. We found that the putative SEC fractions containing exo-AAV1 were significantly more resistant to neutralization by IVIg than conventional AAV1. This increase in resistance was observed over multiple concentrations of IVIG. Despite this increased resistance, at higher concentrations of IVIg, exo-AAV is also neutralized. The mechanism behind this neutralization is currently unknown, but there are several possibilities. First, we have observed in two other studies^6,11 that AAV capsids can be found on both the interior and exterior of EVs. The AAV capsids on the outside may lead to crosslinking of EVs and prevention of transduction at higher antibody concentrations. This may be overcome in the future with better understanding of how to remove or prevent AAV binding to the EV surface, as well as more efficient methods to encapsulate AAV inside EVs. Finally, the neutralization assay using IVIg does not discriminate between anti-AAV capsid neutralization and other mechanisms of neutralization such as antibodies that may have some ability to bind to the EV surface. We asked whether SEC-purified EVs derived from conditioned media of nontransfected 293T could protect standard AAV1 from neutralization (Supplementary Fig. S3). We found that the purified EVs did not protect standard AAV1 from neutralization in contrast to SEC-purified exo-AAV1. This points to a unique property of exo-AAV that is likely mediated by an AAV/EV interaction. Discerning the exact mechanism or mechanisms for this neutralization and neutralization resistance will be an effort of future studies.

One interesting discovery made during our study was the identification of MAAP protein by mass spectrometry which coisolated with the EV-rich, exo-AAV fraction of the iodixanol gradient. MAAP is a recently discovered AAV gene product and was shown to localize to the membrane of transfected cells.²⁰ The discovery of a membrane-associated gene in the relatively small genome of AAV and the observation that its product localizes to membranes and can be found in purified exo-AAV preparations makes it tempting to speculate that AAV may have evolved a mechanism to envelope at least a portion of its capsids in host cell-derived membranes such as EVs. Future studies examining whether deletion of MAAP from AAV genomes reduces, or conversely overexpression enhances, capsid association with EVs will be of great interest.

In the gradient purified exo-AAV1, we noticed a marked faster migration of VP1, VP2, and VP3 bands compared to cell lysate purified AAV1 (Fig. 2b). This could indicate differences in proteolytic processing or posttranslational modifications between enveloped and nonenveloped AAV, although a more detailed mass spectrometry analysis will need to be performed in the future. It will be important to assess any of these differences as it could affect vector potency.

While we recognize that the development of the exo-AAV technology is nascent compared to the mature AAV platform, we are encouraged by the yield of AAV we obtained in the EV-rich fractions, which ranged from ∼40% to 60% for AAV1 and AAV9 serotypes (Table 2). We expect that this will increase as we and others develop the technology and find ways to either package more AAV into EVs or decrease the amount of “naked” AAV that exits the cell.

Overall, our study provides feasibility data of using scalable methods to purify exo-AAV and higher purity preparations of exo-AAV exhibit antibody resistance similar to our prior work with UC-exo-AAV. While additional development of the exo-AAV system remains, such as testing transduction and biodistribution in mice and nonhuman primates and increasing preparation homogeneity, our work here provides an important step toward its clinical utility.

AUTHORs' CONTRIBUTIONS

C.A.M. conceived the study, designed and performed experiments, analyzed the data, and wrote the article. M.C., D.G., and F.E. designed experiments, analyzed the data, and wrote the article. Y.G. and L.D. performed experiments, analyzed the data, and wrote the article. J.N. and C.N. performed experiments.

AUTHOR DISCLOSURE

C.A.M. has financial interest in Chameleon Biosciences, Inc., and Sphere Gene Therapeutics, companies developing an enveloped AAV vector platform technology. C.A.M. also has financial interest in Skylark Bio, which is developing gene therapy to treat hereditary hearing loss. C.A.M.'s interests were reviewed and are managed by Massachusetts General Hospital and Partners HealthCare in accordance with their conflict of interest policies. D.G. is a cofounder, shareholder, chief scientific officer, and managing director of AaviGen GmbH, which is developing AAV vectors for treatment of cardiac diseases.

FUNDING INFORMATION

This work was supported by the National Institutes of Health R01-DC017117 to C.A.M.

SUPPLEMENTARY MATERIAL

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4